Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters

Transcription

1 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters Robert Joumard, M. Rapone, Michel Andre To cite this version: Robert Joumard, M. Rapone, Michel Andre. Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters. Rapport de recherche. 2006, 132p. <hal > HAL Id: hal Submitted on 13 Dec 2010 HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

2 Michel ANDRÉ, Mario RAPONE, Robert JOUMARD Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters ARTEMIS - Assessment and reliability of transport emission models and inventory systems WP3141 research task Report INRETS-LTE 0607 March 2006

3

4 Michel ANDRÉ, Mario RAPONE, Robert JOUMARD Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters ARTEMIS - Assessment and reliability of transport emission models and inventory systems Project funded by the European Commission within the 5 Th Framework Research Programme DG TREN Contract N 1999-RD (ref INRETS: C00-23) With a specific support from the French ADEME agency: ADEME Contract N (ref INRETS: C00-17) Report INRETS-LTE 0607 March 2006

5 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters The authors: Michel André, INRETS tel +33 (0) , fax +33 (0) , michel.andre@inrets.fr Mario Rapone, IM Robert Joumard, INRETS with contribution or help by: Myriam Hugot, INRETS Jean-Marc André, INRETS Peter de Haan, INFRAS Mario Keller, INFRAS The laboratory units: IM: Istituto Motori CNR (National Research Council), viale Marconi Napoli, Italy INRETS: Laboratoire Transports et Environnement, case Bron cedex, France Acknowledgements : These works were possible thanks to: - the financial support by the European Commission and the French ADEME - the availability of the driving data, by TRL (UK), TÜV-Rheinland (Germany), Lab. Of Applied Thermodynamic, Aristotle University of Thessaloniki (Greece) and INRETS (France), in the frame of the DRIVE-modem and BRITE-EURAM Hyzem research projects - the availability of emission data, by TUG (Technical University of Graz, Austria), KTI (Institute for Transport Science, Budapest, Hungary), TNO (Delft, The Netherlands) and INRETS (France), in the frame of the Artemis project, as well as the numerous partners who contributed to the elaboration of the Artemis emission database. - the support of the partners from ARTEMIS 300 who participated at the methodological discussions. INRETS - Lab. Transport and Environment Case Bron cedex France Tel: +33 (0) Fax: +33 (0)

6 Publication data form Publication data form 1 UR (1st author) Transport and Environment Laboratory 2 Project n 3 INRETS report n Report INRETS-LTE Title Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters 5 Subtitle 6 Language E 7 Author(s) Michel ANDRÉ, Mario RAPONE, Robert JOUMARD 8 Affiliation 9 Sponsor, co-editor, name and address 10 Contract, conv. n 12 Notes 11 Publication date March Summary Consequent works have been undertaken within the ARTEMIS project to analyse the influence of the driving cycles as regards the estimation of the emissions. A large number of cycles was reviewed, and their characterization enabled the building-up of a set of contrasted cycles to assess this influence. The emission analyses have demonstrated the significant and even preponderant influence of the driving cycles on the emissions. Quite contrasted emission behaviours were observed for Diesel (rather sensitive to speed and stop parameters) and Petrol cars (rather sensitive to accelerations) The most significant kinematic parameters have been identified. A partial least square hierarchical approach was developed to analyse and estimate the emissions;, which led to a good fit for CO2 but was less or not satisfying for the other pollutants due to a variability between the vehicles and to "high emitting" cars. It was also observed that a model based on the only average speed is unable to predict the emission behaviour induced by the dynamic of the cycles. The question of using dedicated driving cycles for the high- and low-powered cars respectively rather than a unique common set of cycles for all the cars was also raised-up as this procedure aspect induced strong differences in the estimation. The harmonization of the Artemis cars emission database, through a cartography of the cycles has enabled the definition of reference emissions which should enable a better taking into account of the traffic dynamic on the emissions. These results were also implemented for the development of a specific method to compute the emissions at a low spatial scale, i.e. the so-called traffic situation approach. 14 Key Words Speed, acceleration, passenger cars, driving cycle, kinematic parameters, pollutant emission 15 Distribution statement free 16 Nb of pages 136 pages 17 Price F 18 Declassification date 19 Bibliography yes ISBN: INRETS/RR/ ENG 1

7 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters Fiche bibliographique 1 UR (1er auteur) LTE, Laboratoire Transports et Environnement 2 Projet n 3 Rapport INRETS n Report INRETS-LTE Titre Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters 5 Sous-titre 6 Langue E 7 Auteur(s) Michel ANDRÉ, Mario RAPONE, Robert JOUMARD 8 Rattachement ext. 9 Nom adresse financeur, co-éditeur 10 N contrats, conv. 12 Remarques 11 Date de publication March Résumé Des travaux conséquents ont été entrepris dans le cadre du projet de recherche ARTEMIS, pour analyser l incidence des cycles de conduite sur les émissions de polluants des voitures particulières. A grand nombre de cycles a été analysé et leur caractérisation a permis le construction d un jeu de cycles contrastés permettant d analyser cette influence. L analyse des émissions démontre l influence significative et même prépondérante du cycle de conduite sur les émissions. Des comportements contrastés sont observés entre véhicules Diesel (sensibles aux vitesses et arrêts) et voitures essence (plutôt sensibles aux accélérations). Les paramètres cinématiques les plus significatifs ont été identifiés. Une approche hiérarchique basée sur la régression partielle a été développée pour analyser et estimer les émissions. Cette approche est satisfaisante pour le CO2 mais moins pour les autres polluants en raison de la forte variabilité entre les véhicules et en particulier de véhicules «fortement» polluants. Un modèle basé sur la vitesse moyenne se révèle insuffisant à prédire les émissions en raison de la dynamique des cycles. La question de l utilisation de cycles de conduite spécifiques, selon les capacités de motorisation des véhicules plutôt que d utiliser des cycles communs pour tous les véhicules a été soulevée également, compte tenu des distorsions importantes d estimation des émissions induites par cet aspect purement méthodologique. L harmonisation des données d émissions de la base de données Artemis (voitures particulières), au travers d une cartographie des cycles d essai, a permis l élaboration d émissions de référence qui doivent permettre une meilleure prise en compte de la dynamique du trafic sur les émissions. Ces résultats ont été également mis en œuvre pour la construction d une approche spécifique permettant d estimer les émissions à une échelle «locale» (approche dite de situations de trafic). 14 Mots clés Vitesse, accélération, voiture particulière, cycle de conduite, paramètre cinématique, émission de polluants 15 Diffusion libre 16 Nombre de pages 136 pages 17 Prix F 18 Confidentiel jusqu'au 19 Bibliographie oui ISBN: INRETS/RR/ ENG Report INRETS-LTE

8 Summary Summary SUMMARY 3 INTRODUCTION 7 1. REVIEW, CHARACTERIZATION AND SELECTION OF DRIVING CYCLES Methodological aspects Objectives Influence of the driving cycles and their kinematic characteristics Enlarging the coverage of the usual driving cycles Method and principles to derive driving cycles Driving cycles considered Existing cycles The Artemis cycles Driving cycles designed for high and low powered cars Methods Simple approach Speed and acceleration distribution approach Driving cycles and sub-cycles selection Preliminary classification Motorway /main road driving cycles Simple Analysis of the motorway driving cycles Speed - acceleration Analysis for the motorway driving cycles Urban driving cycles Simple analysis of the urban driving cycles Speed - acceleration Analysis for the urban driving cycles Rural-road and sub-urban driving cycles Simple Analysis of the road driving cycles Speed - acceleration Analysis for the rural driving cycles Recapitulation and conclusion EXPERIMENTAL PROTOCOL AND EMISSION DATASETS Adjustments and rules of usage Elaboration of 4 new composite driving cycles Rules of usage and experimental protocol Vehicles tested Other emission data sets The PNR-Ademe emissions dataset The Artemis emissions dataset INFLUENCE OF THE DRIVING CYCLES ON THE EMISSIONS Data sets and method Data sets Coverage of the driving cycles Emissions data and emissions per vehicle Factors influencing the emission Relative importance of the different factors influencing the emissions Identification of the possible level of analysis Influence of the driving type 48 3

9 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters Influence of the motorisation Driving cycles and kinematic parameters influencing emission Remarkable driving cycles Emissions and kinematic parameters Characteristic kinematic parameters regarding the emissions Detailed analysis per cycle Conclusion SENSITIVITY OF THE EMISSIONS TO THE TEST PROTOCOL: COMMON VERSUS SPECIFIC CYCLES Detailed and aggregated comparisons Differences according to driving type Differences regarding vehicle categories Conclusions EMISSIONS MODELLING AS REGARDS KINEMATIC PARAMETERS Possible parameters for an emission modelling Hierarchical approach combining 2 Partial Least Square regression models Case studies Driving cycles Regression models Main results Detailed results Gasoline case studies Detailed results Diesel case studies Conclusion EMISSION DATA HARMONIZATION Emission data in Artemis Artemis data Non-Artemis data Principles of the approach Cartography of the driving cycles Reference cycles Cycles selection for the emission estimation Approaches Analyses and results Emission calculation and refinements Discussion Implications as regards the emissions modelling APPROACH FOR ESTIMATING THE EMISSIONS AT A TRAFFIC SITUATION LEVEL Definition of traffic situations Representative speed data Emission estimation Conclusions 90 CONCLUSIONS 91 Report INRETS-LTE

10 Summary BIBLIOGRAPHY 93 APPENDICES 95 Annex 1. Driving cycles considered for selection (Chapter 1) and harmonization (Chapter 6) 97 Annex 2. Correlation matrix between the kinematic parameters describing the driving cycles 107 Annex 3. Classification of driving cycles as motorway / main roads /rural / urban 109 Annex 4. Experimental protocol 111 A.6.1. First day (or half day) - The ARTEMIS Cycles 111 A.6.2. Second day - Neapolitan D.C. and other ones 111 A.6.3. Third day - Handbook D.C. and last cycle 112 Annex 5. Rules of usage of the cycles 113 A.5.1. Rules of usage 113 A.5.2. Gear box ratio changes 114 Annex 6. Gearshift statistics and test strategy 115 Annex 7. Vehicles tested in the frame of WP Annex 8. The French PNR-Ademe complementary emission dataset 118 Annex 9. Pollutant emissions per driving cycle 119 Annex 10. Classification of the driving cycles from the Artemis emission database 127 Annex 11. Reference emissions according to the driving cycles 129 Annex 12. Reference emissions according to the driving patterns Extrapolations 131 5

11

12 Introduction Introduction These works (task WP3141: influence of the driving cycle of the ARTEMIS 1 research project) were initially and roughly designed to : - review and compare the existing driving cycles as regards their kinematics, representativity and method of determination, - analyse the sensitivity of emissions as regards test cycles and compare emissions data measured with different cycles, - assess the quality of the emissions modelling according to the number and quality of measurement cycles. In this aim, emissions measurements were foreseen on 11 cars using about 16 driving cycles, to be selected amongst the already existing test cycles, or to be adapted or developed. Six laboratories were involved in these tests. In fact, the works conducted around this topic have deviated to a large extent from this very simple experimental design due to the complexity of the question and to the importance of the issues. Indeed: - The driving cycle is the basic material of the emissions measurements and therefore its qualities (representativity, exhaustivity, reproducibility, reliability, etc.) are crucial as regards the quality of the emissions data, factors and modelling. Unfortunately, due to the experimentation costs, the driving cycles are often limited in number (2-3) and in duration (some tenths of minutes), and that probably limits their representativity. In Europe, there are plenty of driving cycles and that should be seen as an advantage. But these cycles were developed in various countries with different methods and assumptions, without coherency between them. Furthermore, there is generally no reliable mean or dataset that would enable the necessary comparisons and getting then the advantage of this cycles multiplicity, and of the large quantity of related emission measurements. - The driving cycle is the only link with the driving conditions or behaviour, i.e. with the actual on-the-road condition. The necessity in ARTEMIS to estimate the emissions as a detailed level (i.e. in one street for a given traffic condition) increased dramatically the need to understand and even to model the link between emissions and kinematic parameters. Although the present task had not such aim, it appeared to be the good framework and at least a good starting point to deal with such a question. - The collection in ARTEMIS of a very large amount of ancient and new emissions data, coming from various laboratories and measured using a large range of driving cycles raised up the question of harmonizing these data and in particular as regards the test cycles. This necessary harmonization should obviously be based on the analyses of these 1 ARTEMIS: Assessment and reliability of transport emission models and inventory systems. Project funded by the European Commission within the 5 th Framework Research Programme, DG TREN. 7

13 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters driving cycles as regards their kinematic content, and on the understanding of the link between driving profiles and emissions. Indeed the very large heterogeneity of the emission data is certainly due to a large extent to the variability between the cycles. Once again and although not foreseen in its frame, the WP3141 task appeared to be a good framework to approach this question. - Although the experimental dataset foreseen in this task enabled to draw significant conclusions as regards the impact of driving cycles on the emissions, it appeared to be generally insufficient to derive practical functions or corrections such as the ones mentioned in the previous point. Complementary emissions dataset were thus considered and analysed. These dataset were much more consequent in size, but unfortunately not so well conceived to measure in a simple way the effect of one or several parameters. Complex data analyses were then required. For these reasons, the scope of this task was strongly enlarged to cover complementary questions such as emissions modelling aspects, emissions data corrections, and by the need of sophisticated analyses of complementary dataset. Finally the works described in that report includes the following main topics: 1. The review and collection of a large range of cycles, their characterisation, and the building-up of a set of contrasted and typical cycles, 2. The tests on chassis dynamometer and measurement of the pollutants emissions of 9 out of 11 cars foreseen (for 2 cars the data was not made available), 3. The analysis of the previous data and of a complementary dataset of 30 vehicles tested using both the Artemis cycles and specific cycles corresponding to their motorization (cycles for the high and low motorization cars), to characterize the influence of the cycles on the emissions, 4. The review of diverse approaches to model emissions as regards the kinematic parameters, and the elaboration of a hierarchical statistic modelling approach using Partial Least Square regression and based on the Artemis database, 5. The analysis of the whole set of hot emissions data for passenger cars collected in the Artemis database and the development of an approach that enabled the harmonization according to the test cycles and the computation of the emissions, 6. The development of a specific method to compute the emissions at a low spatial scale, i.e. the so-called traffic situation approach. Remark: These works led us to deal with a large number of driving cycles, sometimes used or known by other institutions with differing names. The induced confusion led us to a fastidious task in order to harmonize and define clearly each cycle. By convention, in that report a cycle will be named by its family name (giving information on its origin or its context), followed by a cycle name, which attempts to be as explicit as possible. Report INRETS-LTE

14 Review, characterization and selection of driving cycles 1. Review, characterization and selection of driving cycles The first aim of this chapter is the review and collection of a large range of driving cycles. Their description and characterization enable then the selection of a pertinent sample of test cycles, in order to study the cycle influence on emissions. We develop hereafter the methodological aspects of this selection Methodological aspects Objectives These works do not intend to compare different driving cycles as regards their emission level or their quality (which objective would be of quite low interest) but aimed firstly at : 1. analysing the influence of the driving cycle and its "kinematic content" on the pollutant emissions, 2. enlarging the coverage of the emissions test, to driving conditions that are not covered by the cycles used in the others tasks of the Artemis project, 3. possibly, studying the methods and principles used to derive driving cycles. We examine these points as regards their implication on the experimental design and the works to be conducted Influence of the driving cycles and their kinematic characteristics This objective should contribute to a better knowledge of emissions as a function of the driving conditions, and in particular to the improvement of the instantaneous (or microscopic) emissions modelling. The underlying questions are: How to characterise the driving cycles? - Synthetic parameters can be envisaged, such as stop number and duration, maximum and average speeds, acceleration number and level, power or energy of the speed profile. More detailed or complex parameters can also be considered, such as the speed acceleration crossed distribution. Which cycles do we consider? the calculation of the above parameters for each of the available cycles should enable their ranking and the selection of cycles that would be contrasted as regards their kinematic contents (i.e. maximising the variation of the kinematic content for different speed levels). 9

15 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters Enlarging the coverage of the usual driving cycles The selection of cycles should allow getting emission data in ranges that are not covered by the driving cycles used in Artemis (i.e. Artemis cycles, NEDC, FTP), while addressing also the question the influence of driving cycles on emissions. In the following of (De Haan et al., 2001) the following situations are suggested to increase the coverage of the Artemis cycles: - the very congested urban, or stop and go, - stop and go on motorway, - the range of about 100 km/h (motorway with 100 km/h limit), - high (and low) accelerations Method and principles to derive driving cycles Although useful, such a review (as regards the database, the statistical principles and method used, the basis or induced assumptions) is quite difficult as it implies a good documentation on the works, which is not always available. However, several works in that area can already be found (see André 1996, André 2004b, Watson 1995, etc.). On the other hand, it would be useful to analyse the incidence of principles induced by the approaches, and to consider for instance the following aspects: - the "central trend approach" (i.e. cycles are derived to reproduce average or median parameters, etc.), which leads to cover the average driving instead of describing its diversity, - the appropriateness of a unique cycle, independent of the cars performances (high or low motorization, etc.) and of vehicle categories (private or commercial cars, light duty vehicles, etc.), - the homogeneity of driving conditions within a cycle, rather than the heterogeneity observed within real trips, - the design of short sub-cycles within the test cycles, which enrich the accuracy of the measurements and analyses (through more detailed situations) but also decrease the representativity through short measurements (at the sub-cycle scale), - the taking into account of the gear strategy, of the starting conditions, etc., which are certainly important criteria characterizing the cycles Driving cycles considered A large range of driving cycles for passenger cars is available in Europe. A compilation of some of these works can be found in (André, 1996, 2004a and 2004b, Joumard et al. 2000). The list of these cycles, their main characteristics and classification according to the following analyses are provided in Annex 1 (which includes also the other cycles considered in Chapter 6). We provide hereafter a brief description of these cycles, considering first the existing cycles and then those developed in the frame of the Artemis project. Report INRETS-LTE

16 Existing cycles Review, characterization and selection of driving cycles The following driving cycles have been considered in this framework: - Standard (or legislative) driving cycles such as the European NEDC cycles (urban, suburban), the US FTP cycle, the US Highway. - The Inrets cycles: derived from French experimentation on the actual driving, these 10 driving cycles (5 urban, 3 rural, 2 motorway) describe in detail the driving conditions, with an intrinsic homogeneity in speed levels (Joumard et al. 1987). - The Inrets short cycles: derived from the previous ones to measure the cold start effect, they consist in urban and rural short cycles that are repeated numerous times. - The modem cycles: these 14 urban cycles were derived from the monitoring of 60 European cars in actual use (DRIVE-modem research project, André et al. 1994)). Compared to the Inrets cycles and apart from their international character, they introduce the heterogeneity and chronology of the driving conditions within the cycles. - The modem-im cycles and short-cycle: based on the same database and principles than the modem cycles, 4 cycles cover the congested urban driving, the free-urban driving, the rural and motorway driving (André et al. 1998). - The modem-hyzem driving cycles for cars: derived from the DRIVE-modem and BRITE/EURAM-Hyzem research projects, they consist in 2 imbricate sets of 3 (urban, rural, motorway) and 8 cycles (3 urban, 3 rural, 2 motorway), providing two levels of description of the driving conditions (André, 1997). Compared to the modem and modem- IM cycles, they are based on a larger dataset (80 European cars) and on totally revised statistical principles improving their representativity. - The PVU or LDV cycles for 4 categories of light duty vehicles (cars professionally used, light vans, 2.5 and 3.5 tons vans): these four sets of cycles account for 6 to 9 cycles and include urban, rural and motorway empty and loaded driving, as well as specific conditions such as delivery. These cycles were based on a French experimentation with 40 LDV monitored in real-world driving conditions (Joumard et al. 2003). - Short test cycles: 3 short test cycles were considered: the US IM240, the modem-im short-cycle and one short test cycle derived from the European standard cycles (André et al. 1998). - Various other test cycles: New-York City Cycle (NYCC), the non-ftp cycles (US SC03, US06). - The Swiss Handbook driving cycles, developed to cover the Swiss driving conditions in a detailed way (3 cycles, 12 sub-cycles, de Haan et al. 2001) Neapolitan driving patterns recorded in urban and congested contexts The Artemis cycles Based on a large database (called modem-hyzem) on the actual driving in Europe, the Artemis driving cycles have been designed to describe the space of the actual driving conditions in their diversity. Twelve typical driving patterns were identified from the 11

17 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters analysis of the detailed speed and acceleration, and 14 representative sub-cycles were built-up within 3 driving cycles (André, 2004a and b, Figure 1). These test cycles enabled then the measurement of the emissions according to 12 patterns characterizing the European driving, allowing then a detailed analysis of the emissions as regards kinematic parameters. This accuracy in the emissions description was searched for to enable accurate emissions estimation approaches (street level, see Chapter 0). Speed (km/h) 60 urban dense free-flow urban congested, stops congested, low speed flowing, stable times (s) Figure 1: The ARTEMIS urban driving cycle and its structure in driving conditions Driving cycles designed for high and low powered cars These cycles were built-up using the same data and principles than the Artemis cycles, but considering two sub-samples of cars differentiated according to their motorization (power to mass rate). Compared to the Artemis cycles, they introduce then a refinement in the representativity as they are more appropriated to the cars according to their performances (see Andre et al. 2005a), and correspond in a better way to the driving conditions really observed by these two categories of vehicles. Two sets of 5 cycles dedicated to high and low motorized cars (each car being tested according to its characteristics using the one or the other cycles set) have then been developed. The urban, rural, motorway cycles are totally compatible with the Artemis cycles in their structure and method of elaboration. A dense and a free-flow urban cycles are added. Average positive acceleration (m/s2) 1,1 High-powered cars D.C. Low-powered cars D.C. High-powered cars sub-cycles Low-powered cars sub-cycles 0,7 0, Running speed (km/h) Figure 2: Difference in the driving patterns reproduced in the cycles and sub-cycles for high and low powered cars, as regards speed and acceleration. These driving cycles (respectively Artemis.HighMot_motorway, _rural, _urban, _urbdense Report INRETS-LTE

18 Review, characterization and selection of driving cycles and freeurban, and Artemis.LowMot_motorway, _rural, _urban, _urbdense and freeurban) are used respectively by the vehicles with high / low power to mass rate. Diesel vehicles appear to be all in the low category. Due to the similarity between the cycles sets (equivalence in the traffic situation but difference in the driving patterns as highlighted in Figure 2), these cycles constitute a good basis to analyse the influence of the driving patterns on the emissions, even if the differences are very limited (compared to those observed between the large range of cycles). This cycles should also enable to assess the influence of considering a single set of cycles common to all the cars, rather than developing specific cycles according to the vehicle characteristics Methods Considering the above list of cycles and objectives, we attempt to select contrasted cycles for different speed levels, congested cycles (Neapolitan driving, Swiss urban and motorway stopand go cycles), cycles derived according to contrasted approaches (cycles derived according to vehicle performances, homogeneous versus heterogeneous cycles, single cycle compared to a set of contrasted cycles, etc.). Two approaches have been used to characterize and select driving cycles. The first one, a simple approach, is mainly based on the visualization of the driving cycles as a function of 2 kinematic parameters. The second approach is based on the analysis of the kinematical content of the cycles, through the 2-dimensional distribution of the instantaneous speed and acceleration. This approach (which is obviously more rich due to the detailed description of the kinematic) aims at establishing a typology of the test conditions, this typology being then used to select contrasted cycles. In the following analyses, both approaches are presented in parallel and the charts show the combined results (i.e. the typology can be shown on the speed x acceleration charts). This should highlight the interest of the approach and reinforce the conclusions. The driving cycles for light duty vehicles (light vans and vans) were not used for selection as the corresponding driving patterns would not have been appropriated for passenger cars. They are however considered in the analyses for their positioning as regards the other cycles. Some driving cycles are structured in sub-cycles for the emissions analysis. Thus, they can be analysed at the 2 levels: as main cycle and per sub-cycle. Each sub-cycle is then named by the label of the driving cycle followed by the number of the sub-cycle. This is the case for the Swiss Handbook and Neapolitan driving patterns, for the Artemis driving cycles as well as other ones. Some cycles include also pre-conditioning, transitions or post-parts that are generally not considered in the analyses. When the sub-cycle is the appropriate entity to describe a driving condition, only the subcycle is considered as an active observation (i.e. which contribute elaborating the typology), while the whole cycle is considered afterwards as regards this typology (illustrative observation) Simple approach The first approach is mainly based on a simple positioning of the driving cycles as regards 13

19 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters variables that can be considered as relevant for emissions. A high number of variables can be considered: - the average speed (including the stop phase) and the running speed (stop excluded), the first one being less representative of the movement phase, - the speed variation (through the standard deviation, the number of fluctuations, etc.), - the acceleration variation (through the standard deviation of the acceleration, the average positive acceleration, deceleration, the number of accelerations, etc.), - stops number and durations, etc. The correlations between more than 51 of these parameters (calculated for the available driving cycles) are shown in Annex 2. This demonstrates that we should consider several categories of parameters, and within these categories, the parameters are quite well correlated together: - those linked to the size of the cycle (i.e. duration, distance, duration within a certain driving condition, speed range, etc., number of stops, accelerations, etc.), - those describing the speed (average, running, maximum, etc.), - those describing the acceleration (average values, frequencies, etc.). One should also remark that the stop frequency (stop/km) is not correlated to other parameters. The running speed (stops excluded) and the average positive acceleration have been selected for these analyses. Indeed, correlation analyses have shown than these 2 parameters enable a quite satisfying description of the vehicle movement. The average acceleration is quite strongly correlated with other kinematic parameters representing the variability (stops parameters included), while being not too much correlated with the speed parameters. Considering then the field of the driving conditions as regards speed and acceleration, we can select driving cycles allowing a good coverage, and choose cycles that would be well contrasted as regards acceleration for a given level of speed Speed and acceleration distribution approach The second approach is based on the analysis of the 2-dimensionnal distribution of the instantaneous speed and acceleration. This distribution is calculated during the running phases (the stop phases are excluded). Stop duration is taken as supplementary variable. Respectively, 8 and 7 classes have been considered for speed and acceleration. In all, we define then 57 variables (frequencies in % of time) which constitutes an very accurate description of the speed profile. Cycles and sub-cycles are then analysed using the Binary Correspondences Analysis. This enables identification of the principal axes and determining variables. An automatic clustering (classification), allows then the identification of typical and contrasted classes. We have thus an optimal description in classes, enabling a good coverage of the driving conditions. We can then select representative driving cycles in these typical classes. These cycles are obviously contrasted as regards their kinematical content due to the method. This analysis is quite similar to the approach used to derive the Artemis cycles (André 2004a). A slight refinement was however introduced, which concerns the 2-dimensionnal Report INRETS-LTE

20 Review, characterization and selection of driving cycles distribution of the speed and acceleration. Indeed, such matrix includes generally a high number of cases with very low or null figures, the time spent at high speed and high acceleration being very rare. These empty cells are generally a source of trouble with the factorial analysis. The solution adopted for the building-up of the Artemis driving cycles, was to group the situations with low time frequencies (i.e. to reduce the number of acceleration classes at high speed and even to regroup high speed classes). This resulted in a lower accuracy of the description of the driving at high speed. We have here solved this problem through a distortion of the acceleration according to the speed (highlighted in Figure 3). We apply for that a multiplying rate on the acceleration, which is 1 at low speed and increase linearly up to 2 at 140 km/h. This enables artificially an improvement of the description of the high speed driving. Normal 2-dimmensional distribution of speed and acceleration duration in 1/1000, stops excluded Speed (km/h) Acceleration (m/s2) < >140 total < ~ ~ ~ ~ ~ > total Distribution with a distortion of the acceleration duration in 1/1000, stops excluded Speed (km/h) Acceleration (m/s2) < >140 total < ~ ~ ~ ~ ~ > total Figure 3: Two-dimensional distribution of the speed and acceleration obtained with the set of driving cycles (top) and distortion introduced to improve the description of the driving at high speed (down the acceleration classes are only indicative in that case) Driving cycles and sub-cycles selection The selection of cycles / sub-cycles will be firstly based on the second approach, because this one is the most statistically significant and it offers also criteria of ranking and representativity of the cycles. The first approach - more arbitrary will be considered to validate the choices. Furthermore, to simplify the experimental procedure, an optimization of the selection should: - favour entire set of sub-cycles (within a cycle) and of cycles (belonging to the same family), - select in particular the Artemis sub-cycles as they constitute a reference for most of the emissions measurements done in the Artemis project and in other related national projects Preliminary classification The 2 above approaches cannot be conducted considering directly the whole set of driving cycles, because of the too large field covered. Furthermore, preliminary analyses have demonstrated that considering the whole range of driving doesn t enable a pertinent description 15

21 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters of the different driving types (at least urban and motorway) and lead to trivial conclusions such as opposing high to low speeds. It is therefore pertinent and useful to get a first classification, allowing analyses by rough classes of cycles. There are several possibilities for this preliminary classification: - to consider urban, rural and motorway cycles according to their definition or official status ; this approach appeared as not appropriate as definitions are not always coherent. The following analyses will demonstrate it. - to define large classes according to the speed level (for instance, 0-40 km/h for urban driving cycles, km/h for road driving, km/h for main road driving, km/h for motorway driving). Although quite arbitrary, such a breakdown enables the identification of quite homogeneous groups of cycles. The analysis of the speed and acceleration distribution (as described above) was used here. This method identifies the factorial axes as follows: the Axis 1 is typically a speed axis, which opposes high speeds to low acceleration rates and frequency. The Axes 2 and 3 represents well the frequency of the accelerations, and differ as regards their speed range. The classification (full list in Annex 3) enables easily the identification 4 homogeneous sets of driving cycles: - 14 motorway cycles (or sub-cycles) driven at an average speed of 121 km/h, with very low occurrence of stops. The most representative of this group are the Artemis motorway sub-cycles and their equivalent within the High- Low-powered cycles, the Handbook.R1_I cycles corresponding to the driving on main roads, at a quite stabilized speed of 92 km/h on average, with one stop every 14 km. The Artemis rural_4 sub-cycles and equivalent, LDV-PVU.CommercialCars.motorway_1, Handbook.R1_II, Inrets.autoroute1 and 2, US06, modemhyzem.road2 and motorway1 belong to this group. It is interesting to note that numerous motorway and rural cycles are part of this group cycles corresponding to a driving on rural roads, possibly also in sub-urban or extraurban conditions (as suggested by the affectation in this class of one Artemis urban subcycle, several free-flow urban cycles or sub-cycles, the European standard extra-urban cycle, and even several Neapolitan driving patterns). These cycles have together an average speed of 50 km/h, and record about one stop every 2 kilometres (which gives a rough idea of length between intersections). The most significant cycles of this group are Inrets.route2, modemhyzem.road and road1, modemim.road, as well as Napoli.17 and 20 patterns, US.SC03, modemim.short. We note that the FTP first phase and whole cycle, the US_Highway, IM240, and European EUDC belong also to this group urban cycles / sub-cycles, with an average speed of 15 km/h, 30% of the duration at stop, and about one stop every 200 meters, which typically corresponds to an urban area. The most representative cycles of this category are respectively Artemis.urban1 and 4, US.NYCC, modem.urban1 and 8, modemhyzem.urban2, Inrets.urbainfluide1, modemim.urban_slow. The FTP second phase, US. NYCC, European ECE belong to this category, which includes most of the free-flow and congested cycles and sub-cycles and most of the Neapolitan patterns. Report INRETS-LTE

22 Review, characterization and selection of driving cycles. The classification process as well as the high variability of the cycles are well illustrated in Figure 5 and Figure 6. The average characteristics (computed per group of cycles) are provided in Table 1. Class Number of cycles or sub-cycles Average speed (km/h) Running speed (km/h) Average positive Number of acceleration accelerations (m/s2) /km Stop duration (%) Number of stops /km 1 motorway cycles ,41 0,4 0,3 0,02 2 main roads (highways) cycles 3 rural (and suburban) cycles ,60 1,1 1,5 0, ,68 2,7 6,9 0,5 4 urban cycles ,73 8,1 29,8 5,8 All together ,67 2,6 16,1 1,1 Table 1: Average characteristics of the clusters of driving cycles, determined through analysis and automatic classification of the driving cycles described by their crossed distribution of the instantaneous speed and acceleration (average values calculated for the samples) As the motorway and highways groups are quite comparable and relatively limited in number, they will be analysed together in the following. 17

23 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters Report INRETS-LTE

24 Review, characterization and selection of driving cycles Average positive acceleration (m/s2) 1,1 0,9 0,7 0,5 Handbook.R4_III 0,3 Artemis.urban_4 Handbook.R4_II Artemis.urban_2 Artemis.urban_1 Legislative.ECE_2000 Artemis.urban_3 Artemis.urban Handbook.R4_I Artemis.rural_3 Handbook.R3_I Handbook.R3_II WP Selection of driving cycles Artemis.urban_5 Legislative.US_FTP2 Legislative.US_FTP Handbook.R3_III Legislative.US_FTP1 Artemis.rural_1 Legislative.NEDC_2000 Artemis.rural Artemis.rural_2 Handbook.R2_III Legislative.EUDC Legislative.US06 Artemis.rural_4 Legislative.US_HWAY Handbook.R2_II Handbook.R2_I Artemis.rural_5 Handbook.R1_III Handbook.R1_II Artemis.motorway_150_2 Handbook.R1_I Artemis.motorway_150 Artemis.motorway_150_1 Artemis.motorway_150_3 Artemis.motorway_150_ Running speed (km/h) Figure 4: Visualization of the main cycles as regards speed and acceleration Artemis cycles Swiss Handbook cycles Neapolitan driving patterns Legislative cycles High motorisation cycles Low motorisation cycles Other cycles 19

25 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters Average positive acceleration (m/s2) WP Urban driving cycles 1,1 Artemis.urban_3 0,9 Artemis.urban_4 Artemis.urban Artemis.urban_5 Artemis.urban_1 Legislative.US_FTP2 0,7 Handbook.R3_III Legislative.ECE_2000 Handbook.R4_III Legislative.NEDC_2000 0,5 0,3 Handbook.R4_II Artemis cycles Swiss Handbook cycles Neapolitan driving patterns Legislatives cycles Others cycles Running speed (km/h) Figure 5: Visualization of the urban cycles as regards speed and acceleration the urban class results from the automatic classification according to the 2-dimensional distribution of the instantaneous speed and acceleration Report INRETS-LTE

26 Average positive acceleration (m/s2) WP Suburban/rural, main road and motorway driving cycles Review, characterization and selection of driving cycles 0,9 0,7 Handbook.R4_I 0,5 0,3 Artemis.urban_2 Artemis.rural_3 Legislative.US_FTP Handbook.R3_I Handbook.R3_II Legislative.US_FTP1 Artemis.rural_1 Handbook.R2_III Artemis.rural Legislative.EUDC Artemis.rural_2 Legislative.US_HWAY Legislative.US06 (83km/h, 1,3m/s2) Artemis.rural_4 Handbook.R2_II Handbook.R2_I Artemis.rural_5 Artemis.motorway_150_2 Handbook.R1_III Handbook.R1_II Motorway cycles (class 1) Main road cycles (class 2) Sub-urban and Rural cycles (class 3) Artemis3 Artemis.motorway_150 Handbook.R1_I Running speed (km/h) Artemis.motorway_150_1 Artemis.motorway_150_3 Artemis.motorway_150_4 Figure 6: Visualization of the rural / suburban, main road and motorway cycles as regards speed and acceleration the 3 classes result from the automatic classification according to the 2-dimensional distribution of the instantaneous speed and acceleration 21

27 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters Report INRETS-LTE

28 Review, characterization and selection of driving cycles 1.5. Motorway /main road driving cycles Simple Analysis of the motorway driving cycles Average positive acceleration (m/s2) 0,7 0,5 0,3 Artemis.HighMot_rural_4 Inrets.autoroute1 modemhyzem.road2 Legislative.US06 Artemis.LowMot_rural_4 LDV_PVU.CommercialCars.motorway_1 Artemis.rural_4 Artemis.rural_5 WP Main road and motorway driving cycles Artemis.HighMot_motorway_2 Inrets.autoroute2 Handbook.R2_I Artemis.LowMot_rural_5 Artemis.HighMot_rural_5 Handbook.R1_III 23 modemim.motorway modemhyzem.motorway Artemis.motorway_150_2 Artemis.LowMot_motorway_2 modemhyzem.motorway1 Handbook.R1_II Artemis.motorway_130_3 LDV_PVU.CommercialCars.motorway_2 Artemis.motorway_150_1 Artemis.HighMot_motorway_3 Handbook.R1_I Artemis.LowMot_motorway_1 Artemis.motorway_130_4 Artemis.motorway_150_4 Artemis.LowMot_motorway_3 Artemis.HighMot_motorway_1 Artemis.motorway_150_3 Artemis.HighMot_motorway_4 Artemis.LowMot_motorway_ Running speed (km/h) Figure 7: visualization of the motorway and highway cycles as regards the running speed (in x) and the average positive acceleration (in y) the classes 1-8 results from the automatic classification based on the 2-dimensional distribution of the speed and acceleration (see next section) The simple analysis of the driving cycles as a function of running speed and acceleration (Figure 7) suggests: - the Artemis.motorway cycle enables a good coverage for high speeds (sub-cycles 1, 3, 4) and high acceleration (sub-cycles 1, 4). The Artemis.motorway130 cycle (limited to 130 km/h) does not offer such a contrast (except for sub-cycle 1) - the two sub-cycles with high acceleration can be opposed to the corresponding sub-cycles from the motorway cycles dedicated to vehicles with low power-to-mass rates (Artemis.LowMot_ cycles). - Generally and for high speeds, both Artemis.motorway and Artemis.motorway130 cycles show a high contrast with Handbook.R1_I (which has lower acceleration rate). LDV- PVU.CommercialCars.motorway_1 (dedicated to vehicles with professional use) is also well and directly opposed to Handbook.R1_I. - Handbook driving patterns R1_I, II, III and R2_I have relatively low acceleration rates. 101 cycles 102 cycles 103 cycles 104 cycles 105 cycles 106 cycles 107 cycles 108 cycles All

29 Analysis of the cars pollutant emissions as regards driving cycles and kinematic parameters It is then interesting to use these cycles as representative of driving patterns with low acceleration, opposed to the Artemis motorway sub-cycles. Considering the lowest speed levels, we observe : - at about 105 km/h, a strong contrast between Handbook.R1_I and II on one side and modemhyzem.motorway1 or modemim.motorway (high acceleration). - at about 90 km/h, a contrast in acceleration between Artemis.rural_5 and Handbook.R2_I on one side (with quite low acceleration level) and Artemis.rural_4, or worse Artemis.HighMot_rural_4 (vehicles with high power-to-mass rate) or LDV- PVU.CommercialCars.motorway_1 (dedicated to vehicles with professional use) that presents high acceleration rate. We should also note the very high level of acceleration of the US cycle US06 (dedicated to cover high speed and acceleration not covered in FTP75). This cycle is not represented because it is out of the scale. It won t be used in the following as we are not well aware about its context. This simple analysis suggests then the use of the following cycles to get a contrasted coverage of the driving conditions: - Artemis.motorway (alternatively Artemis.motorway130), Artemis.LowMot_motorway (reaching also 150 km/h) when possible, Handbook R1 (sub-cycles 1 to 3) and R2 (sub-cycle 1) for high speeds, - modem-hyzem.motorway or modemim.motorway, Artemis.rural at about 105 km/h, - Artemis.HighMot_rural, LDV-PVU.CommercialCars.motorway_1 at about 90 km/h Speed - acceleration Analysis for the motorway driving cycles The classification of the driving cycles enables the identification of 8 contrasted classes described in Table 2, and shown in the principal axes in Figure 8. Roughly, the axis 1 is linked to the speed level, the second axis is linked to the stop occurrence, and the third one is linked to the acceleration level. We observe a quite large contrast for high speeds. Report INRETS-LTE